[Hardik K (hardikkaurani)]

Motivation and Context

Fixes #14617

In the Python MistralAIChatCompletion connector, metadata["usage"] was being set to a 1-tuple (CompletionUsage(...),) instead of CompletionUsage(...) due to a trailing comma in mistral_ai_chat_completion.py.

Because of this:

1. The telemetry decorator checks isinstance(usage, CompletionUsage), which evaluates to False for a tuple, silently omitting gen_ai.usage.input_tokens and gen_ai.usage.output_tokens from telemetry spans.
2. Callers accessing result.metadata["usage"].prompt_tokens encounter AttributeError: 'tuple' object has no attribute 'prompt_tokens'.

Description

• Removed trailing comma in _get_metadata_from_response within MistralAIChatCompletion so metadata["usage"] is assigned CompletionUsage.
• Added assertions in test_get_chat_message_contents verifying that metadata["usage"] is an instance of CompletionUsage with expected token counts.

Fixes #14617

cc Vincent Biret (@baywet) Eduard van Valkenburg (@eavanvalkenburg)

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[Hardik K (hardikkaurani)]

Hi Vincent Biret (@baywet) Eduard van Valkenburg (@eavanvalkenburg), opened this PR to fix #14617 where MistralAIChatCompletion was setting metadata["usage"] to a tuple due to a trailing comma, breaking telemetry spans and caller token access. Regression assertions included in test_mistralai_chat_completion.py. Ready for review, thank you!